Re: request for help: huge score

[David Kastrup]

Thomas Morley <address@hidden> writes:

> 2016-11-23 11:30 GMT+01:00 Thomas Morley <address@hidden>:
>> Hi all,
>>
>> currently there's some ongoing heavy work to make LilyPond work with guilev2.
>>
>> I'd like to test the current state with a huge score, something like
>> an entire sinfony or the like.
>> I don't have a huge score of this kind at hand.
>> May I ask if somebody could send me one?
>> It should compile with a recent devel-version to ensure comparabilty.
>>
>> A huge score may be too much for an attachment on the list.
>> Please try zipped or a private mail.
>>
>> Thanks,
>>   Harm
>
> Many thanks for all the links and files.
>
> Currently I'm waiting for a doc-build to get finished.
> In the 5th hour now, this is getting insane ...

At least it hasn't crashed yet.

This is only partly our doc build process I think.  Of course I don't
really know just how slow your computer actually is.  I think a main
holdup may be Guilev2, likely partly due to our current code base not
being able to work with precompiled .go files, and partly because the
Guile developers are of the opinion that the performance of interpreted
code does not matter since "everything should be compiled anyway".
Which we of course cannot do for embedded Scheme that is encountered
ad-hoc in the files.

The performance problem will likely be quite resilient, and once
dependable numbers for comparison have been collected, it might make
sense to contact the Guile developer list with the findings and ask for
help finding remedies.

The tricky thing will be engaging them to do provide more meaningful
help than "you must be doing it all wrong".

We are practised in pounding the table demanding minimal examples and
reproducible recipes: of course providing such to the Guile developers
will be helpful as well, and for that a dependable Lilydev might be a
good anchoring point for such a recipe.

-- 
David Kastrup